1
Enabling Path of Digital Pathology to
Personalized Medicine
Session #74, February 12, 2019
Matthew G. Hanna, MD, Memorial Sloan Kettering Cancer Center
Rajendra Singh, MD, The Mount Sinai Hospital
2
Matthew Hanna, MD
Advisory Board, PathPresenter
Rajendra Singh, MD
Founder, PathPresenter
Conflict of Interest
Have no real or apparent conflicts of interest to report
3
Agenda
Intro
Ecosystem
Educational
Initiatives
Current
Landscape
end
Q/A
Digital
Pathology
PathPresenter
FUTURE
AI/ML
Personalized
medicine
4
Learning Objectives
Define what is digital pathology and what is its promise
Design an effective program that will train the pathologists of today and
tomorrow in digital pathology
Practice using digital pathology for educational, research, and clinical
purposes
Outline path of pathology to digital pathology to computational pathology
and finally to personalized medicine for patients
5
6
Film photography is
now a niche field
2001
E-book market jumpstarted in 2009.
Did not invest in digital solutions.
2011
Failed to identify digital
photography as a disruptive
technology, had to sell film,
patents, scanners to stay alive
2012
Failed to innovate a digital solution.
Netflix added streaming services in 2010.
2010
Years Companies Filed for Chapter 11
7
9
Life of Patients’ Specimens
10
Digital Pathology Basics
11
THE FIVE RIGHTs OF DIGITAL PATHOLOGY
RIGHT PATIENT
RIGHT SLIDE
RIGHT
PATHOLOGIST
RIGHT TIME
“Telepathology”
RIGHT
DIAGNOSIS
12
Digital Pathology Ecosystem
Resource: Hanna, M. G., & Pantanowitz, L. (2019). Digital Pathology. In R. Narayan
(Ed.), Encyclopedia of Biomedical Engineering, vol. 2, pp. 524532. Elsevier.
13
Digital Pathology Subsystems
Software
Hardware
Whole slide image
Slide Scanner
Slide scanning
Robotics/slide handling
Optics/Lighting
Tissue detection
WSI viewer
File format
WSI acquisition
WSI repository
Compression
Image analysis
14
Digital Pathology Process
ARMS
Acquisition
Retrieval/Storage
Manipulation
Sharing
15
1x magnification
20x magnification
40x magnification
16
Digital Pathology Benefits
17
On-demand Archive/Case management
18
Glass slides storage
Move slides from local
To remote
↓ Costs
19
Archival glass slide requisitions had a 93% decrease in requests
47,387
186,607
424,901
19,369
20,745
12,336
1,426
-
5,000
10,000
15,000
20,000
25,000
-
50,000
100,000
150,000
200,000
250,000
300,000
350,000
400,000
450,000
2014 2015 2016 2017
Glass Slide Requests (n)
Digital Slides (n)
Digital slides available Glass slide requests
20
Slide storage costs projected savings of $274,000/year
Due to decreased vendor services
(i.e. asset retrieval, storage proximity, labor)
21
Image Analysis
22
Wang, Khosla, … Beck (2016) https://arxiv.org/abs/1606.05718 Camelyon16 (JAMA, 2017)
Machine Learning
23
Big Data meets Pathology
Massive volume of digital data generated from WSI & bioinformatics/molecular data
Critical for personalized medicine, health systems, basic research and “Big data”
Dataset sizes: Computer vision vs. computational pathology
1 whole slide = 100 X 60,000 = 6 billion pixelsFrom Fuchs, 2017
Computational Pathology
24
Why is this possible now?
Image datasets
Cognitive
algorithms
Fast
computing/GPUs
25
Challenges
Not always being done with pathologist involvement, mostly being
done in computer science departments
No universal platform to aggregate and share the vast amount of
data generated by pathology across thousands of hospitals, medical
centers and reference laboratories
Available solutions are often scanner specific, lack useful
apps/tools for Pathologists, lack active participation of
pathologists and lack high quality aggregated data
26
You would feel comfortable providing primary diagnosis using
digital pathology, with retrieval of glass slides available upon
request
You would feel comfortable providing primary diagnosis using
digital pathology, without availability of glass slides
27
What can be accomplished?
Education
Trainee/faculty education/CME
Clinical
Create efficiencies for pathologists
Research
Identify digital prognostic markers
28
Education
29
Why is there a need?
Lack of open-access to digital slides or are
restricted by firewalls to download software
Learning from digital slides is not only having
access to digital slides; it should also simulate
current teaching techniques
30
CURRENT TEACHING TECHNIQUE
Outdated, Unscalable, Limited
http://www.teachingmicroscopes.com
31
Current landscape of Digital
Educational Resources
PathXL
32
DPA
33
CAP Case of the Month
34
PathXL
http://www.pathxl.com/pathology-education-tutor
35
PathPresenter
Digital pathology company built by pathologists for pathologists
Focused on building software to enhance and standardize the
learning and teaching experience in pathology
Provide a multidisciplinary educational environment
36
Current PathPresenter Apps for education
Developed Applications for Public or Institutional Use
MySlideBox
Manage your digital slides and
folders.
High Yield
Learn and study from hand-picked
cases to prepare for board
examinations, or to brush up on must-
know diagnoses
My Presentations
Create and manage your presentations
using our extensive Slide library or by
uploading your own slides
Quiz
Search, view, or share from thousands
of cases covering all medical
subspecialties
Slide Library
Search, view, or share from thousands
of cases covering all medical
subspecialties
Group Chat
Create group with other members and
discuss slides.
Analysis
Create Analysis and
collaborate to add
annotations and discussions
in Groups
QA
Cross check the Quality of
Diagnosis by assigning to the
expert reviewers
37
MySlideBox
38
Slide Library
39
My Presentations
40
High Yield Sections
41
Quiz Module
42
Steps to Facilitate Education
Provide an easy to use platform for trainees, faculty,
and pathology departments
Provide high quality content
Content validated by experts
Publicized on social media
43
Path of Pathology to Personalized Medicine
Bring digital
pathology to
the world
Provide a new
standard medical
presentation
platform
Make
pathologist
comfortable
using digital
pathology
High quality
curated
digital
images
Pathology to
personalized
medicine
Growth in
pCAD
44
Use Cases
Physicians
Ready access to a wealth of data and images, with seamless
medical presentations
Institutions
Branded platforms with site specific content, monetized to the
medical community
Pharma
Agnostic platform coupled with pathology data and AI apps for
real time analysis
LEARN, TEACH, CONSULT, RESEARCH, PUBLISH, CME
45
Is it really working?
Usage by country/continent
>162
Countries
Users
>62,000
>10,000
Digital Images Apps
8
46
30%
Why this will only increase
Future
Pathologists
Number of
cancer cases
70%
47
How will AI affect Pathology
Education
De-skilling
Access to resources
48
Clinical & Research
49
Digital Pathology &
Personalized Medicine
50
51
Digital Pathology Slide
52
Cancer
registry
data
53
54
55
Clustering corresponded with molecular gene expressions and prognosis
56
Survival and Relapse
57
58
https://ai.googleblog.com/2017/03/assisting-pathologists-in-detecting.html
59
https://ai.googleblog.com/2018/04/an-augmented-reality-microscope.html
60
Assistive screening tools for pathologists
Resource:
https://ai.googleblog.com/2018/04/an-augmented-reality-microscope.html
https://ai.googleblog.com/2018/10/applying-deep-learning-to-metastatic.html
LYNA (LYmph Node Assistant)
ARM
(Augmented Reality Microscope)
61
Novel Technologies: Multiplex
62
63
(h) autofluorescence
(i) Red = tumor
Green = stroma
(j) Double positivity
yellow = HER2+/Ki-
67+
blue = HER2/Ki-67
red = HER2/Ki-67+
green = HER2+/Ki-67
Multispectral Imaging. A Review of Its Technical Aspects and
Applications in Anatomic Pathology. J. R. Mansfield
HER2 PR Ki67 ER
64
Will AI/New tech Lead to Better Patient Care?
65
Digital pathology is increasingly being used and will be a
key enabler of personalized medicine.
Many areas of medicine including clinical, education, and
research are creating and utilizing digital imaging
applications including AI/machine learning.
Pathologists’ current routines is transforming and being
impacted by digital pathology.
We need to continue to partner with industry vendors and
continue to build applications of WSI for clinical use
including primary diagnosis and image analysis to help
drive personalized medicine.
Take Home Messages
66
The Road to Personalized Medicine
67
Matthew G Hanna, MD
mghannamd@gmail.com
Rajendra Singh, MD
skinpathology@gmail.com
Please remember to complete online session evaluation
Questions